Properly Rebooting a Datastore
Posted: Fri Jul 26, 2013 2:44 pm
I'm in a WMware environment and a couple time a year I patch my datastores (put in windows patches, update Starwind, update RAID card firmware/drivers, update NIC Drivers, etc). So I'm wondering what is the proper procedure for do this?
Here's what happened last time:
[This was on a NON HA datastore]
Inside of VCenter I shut down all the VMs on all the iSCSI targets on the datastore I wanted to patch. I then went through and applied updates and patches where needed. It took several reboots. Once done, I had all the VMHosts rescan the datastores, making sure all hosts were seeing all iSCSI targets again. They were, but there were a dozen or so VMs that the VCenter decided were "Inaccessible". These were scatter across several iSCSI targets on the datastore I patched (so 2 of the 14 VMs on a particular target were seen as "Inaccessible"). They were scatter across several VMHosts. There was no pattern. So I had to remove the VM, browse the datastore and readd the VM. And when I power the VM back on I then had to answer whether the VM was copied or moved. This was a very tedious and time consuming processes that I don't think should have happened. I'm not suggesting this was a Starwind issue/bug/error, but I'm hoping someone here can point out the step(s) I missed to properly down a Starwind datastore without losing VM in my VMWare cluster.
Thanks,
Here's what happened last time:
[This was on a NON HA datastore]
Inside of VCenter I shut down all the VMs on all the iSCSI targets on the datastore I wanted to patch. I then went through and applied updates and patches where needed. It took several reboots. Once done, I had all the VMHosts rescan the datastores, making sure all hosts were seeing all iSCSI targets again. They were, but there were a dozen or so VMs that the VCenter decided were "Inaccessible". These were scatter across several iSCSI targets on the datastore I patched (so 2 of the 14 VMs on a particular target were seen as "Inaccessible"). They were scatter across several VMHosts. There was no pattern. So I had to remove the VM, browse the datastore and readd the VM. And when I power the VM back on I then had to answer whether the VM was copied or moved. This was a very tedious and time consuming processes that I don't think should have happened. I'm not suggesting this was a Starwind issue/bug/error, but I'm hoping someone here can point out the step(s) I missed to properly down a Starwind datastore without losing VM in my VMWare cluster.
Thanks,